Silent refactor broke CI validation mode while all tests kept passing

·1 views

A developer discovered a logic bug introduced during a routine code restructuring of a deployment script's validation mode. Moving an early-return check above the workspace-building function meant the validation stage exited before actually authenticating, resolving, or parsing anything. Because no test explicitly asserted that validation performs real checks — only that it skips publishing — the entire test suite continued to pass undetected. The bug was caught only by manually reading the code diff, not by any automated test. The fix restored the original execution order and added an explanatory comment warning future developers not to move the line.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

druvu-lib-web 2.0 lets Java developers write web templates in PHP syntax

druvu-lib-web 2.0.0 is a Java web framework built on Jetty 12 that uses PHP as its templating language. The engine supports real PHP 8 syntax — including variables, control structures, heredocs, and 115 standard functions — without requiring an actual PHP runtime. All template logic is executed in Java, and server-side attributes are passed directly as PHP variables in the page. A key change from version 1.x is that HTML output is escaped by default, reducing XSS risks unless developers explicitly opt out using raw(). The project is open source, with code, documentation, and a runnable example available on GitHub.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds Hindi AI Voice Agent in 10 Days to Shield Families from Scams

A developer built Suraksha Saathi, an AI voice agent designed to help Indian families identify scams, as part of the VoiceForBharat 10 Days of AI Voice Agents Challenge. The tool allows users to converse naturally in Hindi to verify suspicious messages, such as fake lottery notifications on WhatsApp. It was built using Murf Falcon for text-to-speech, LiveKit for real-time voice, and Google Gemini as the underlying language model. A key technical hurdle involved database lookups before conversations began, which was resolved by separating that logic into a standalone Python script rather than relying on AI prompts. The project went from an empty template to a fully functional anti-fraud agent within ten days and has been shared publicly on GitHub.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Kubernetes Services Explained: How Stable Virtual IPs Replace Unreliable Pod IPs

In Kubernetes, Pod IP addresses are temporary and can change whenever a Pod crashes or is replaced, making direct Pod-to-Pod communication unreliable. Kubernetes Services solve this by providing a stable virtual IP, called a ClusterIP, that acts as a consistent front door for a group of Pods. A Service uses label selectors to identify which Pods should receive traffic, routing requests without the client needing to know individual Pod IPs. The Service's port field defines what the client connects to, while targetPort specifies the actual port on the destination Pod. This abstraction allows Pods to be created, destroyed, or replaced freely without disrupting the applications that depend on them.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Google-MIT Study: Multi-Agent AI Can Cut Performance by 70% Without Right Design

A joint study by Google Research and MIT tested 180 configurations of multi-agent AI systems across three model families and five network architectures, keeping prompts, tools, and compute budgets identical while only changing how agents were connected. Researchers found that performance swung dramatically — from an 81 percent improvement to a 70 percent drop — depending solely on the network topology used. For parallelizable tasks, a centralized coordination structure outperformed a single agent by over 80 percent, while sequential reasoning tasks saw every multi-agent variant degrade performance by 39 to 70 percent. Averaged across all experiments, multi-agent systems delivered virtually no overall gain compared to a single agent, at just plus 0.2 percent. The findings suggest that adding more AI agents without deliberate architectural planning can be counterproductive, and that connection design matters more than model choice or agent count.

0 comments Read more at DEV Community